updates for compatibility with recent/pending core EpiModel changes

[chad-klumb]

closes #34

Some notes:

• since summary_nets is a separate module, it must be included in the module.order argument (in the EpiModelCOVID tests); this requirement could be eliminated by instead calling summary_nets at the end of the resimulation module (where nwstats were formerly obtained); since there is already an issue for moving nwstats to the prevalence module (move the nwstats calculation to the prevalence module EpiModel#811) I'm not planning to move it into the resimulation module

• EpiModelCOVID initialization modules now perform network resimulation at time 1, consistent with core EpiModel

• vertex attributes are now copied to dat using copy_nwattr_to_datattr (within init_nets), which differs slightly in what vertex attributes it excludes as compared to the previous copying mechanism in EpiModelCOVID initialization

[chad-klumb]

@smjenness some questions related to this:

• Is there any significance to the edges in networks 4:6 before the lockdown time in the cruise ship model? Likewise for networks 1:3 after the lockdown time. (They appear to be completely unused.) If not, I think it would make more sense to remove them, especially as relates to e.g. nwstats output.

• Will the cruise ship models always have duration 1? (This is the case in the tests.)

• In general, are networks 1 and 4 allowed to have different model terms, coefficients, constraints, etc.? (In the tests they are the same.) Likewise for networks 2 and 5, 3 and 6.

[chad-klumb]

Note that this PR includes updates for netsim_dat class, depart_nodes, and arrive_nodes, corresponding to EpiModel/EpiModel PRs #849, #845, and #855.

[smjenness]

Thanks!